Typically, the test type exams with binary answers per question of the type:true / false are corrected granting a +1 to the correct answers and a -1 tothose that are not. This mainly obeys to a purely statistic criterion. From our point of view this method of qualication is not suitable and therefore should be amended as soon as possible.In this work we propose two alternative methods that seek to favor goodstudents to the detriment of those who tend to speculate by randomly answering questions of which they do not know the answer. The two methods we propose benefit students who do not answer questions randomly. The second one does not harm them with respect to the classical method of granting +1 per correct answer and -1 per incorrect answer.The proposals made extend to multiple choice exams with an arbitrary and xed number of answers for each question with a single answer being correctper question.